Implementation:Run llama Llama index Index As Query Engine

Knowledge Sources	LlamaIndex LlamaIndex Query Engine
Domains	RAG, LLM_Integration
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete method on BaseIndex that composes a retriever and response synthesizer into a RetrieverQueryEngine for querying indexed data.

Description

The as_query_engine method is the primary entry point for creating query engines from any LlamaIndex index. It delegates to the index's internal as_retriever method to obtain a retriever, then constructs a RetrieverQueryEngine via its from_args classmethod. This classmethod wires together the retriever, an LLM, a response synthesizer (governed by response_mode), optional node postprocessors, and prompt templates into a fully configured query pipeline.

Usage

Call as_query_engine() on any index instance (e.g., VectorStoreIndex, SummaryIndex, KeywordTableIndex) after the index has been built or loaded from storage. Pass configuration kwargs to control retrieval behavior and response synthesis strategy.

Code Reference

Source Location

• Repository: run-llama/llama_index
• File: llama-index-core/llama_index/core/indices/base.py
• Lines: L491-516 (as_query_engine method)
• File: llama-index-core/llama_index/core/query_engine/retriever_query_engine.py
• Lines: L37-128 (RetrieverQueryEngine class and from_args)

Signature

# BaseIndex.as_query_engine
class BaseIndex(Generic[IS]):
    def as_query_engine(
        self,
        llm: Optional[LLMType] = None,
        **kwargs,
    ) -> BaseQueryEngine:
        # Resolves retriever via self.as_retriever(**kwargs)
        # Delegates to RetrieverQueryEngine.from_args(retriever, llm=llm, **kwargs)
        ...

# RetrieverQueryEngine.from_args
class RetrieverQueryEngine(BaseQueryEngine):
    @classmethod
    def from_args(
        cls,
        retriever: BaseRetriever,
        llm: Optional[LLM] = None,
        response_synthesizer: Optional[BaseSynthesizer] = None,
        node_postprocessors: Optional[List[BaseNodePostprocessor]] = None,
        response_mode: ResponseMode = ResponseMode.COMPACT,
        text_qa_template: Optional[BasePromptTemplate] = None,
        refine_template: Optional[BasePromptTemplate] = None,
        summary_template: Optional[BasePromptTemplate] = None,
        output_cls: Optional[BaseModel] = None,
        use_async: bool = False,
        streaming: bool = False,
        verbose: bool = False,
        **kwargs,
    ) -> "RetrieverQueryEngine":
        ...

Import

# as_query_engine is a method on any index instance
from llama_index.core import VectorStoreIndex

index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()

I/O Contract

Inputs (as_query_engine)

Name	Type	Required	Description
llm	LLMType or None	No	LLM to use for response synthesis; defaults to Settings.llm
**kwargs	dict	No	Forwarded to both as_retriever() and RetrieverQueryEngine.from_args()

Inputs (RetrieverQueryEngine.from_args)

Name	Type	Required	Description
retriever	BaseRetriever	Yes	Retriever that fetches relevant nodes from the index
llm	LLM or None	No	Language model for synthesis; resolved from Settings if omitted
response_synthesizer	BaseSynthesizer or None	No	Pre-built synthesizer; if provided, overrides response_mode
node_postprocessors	List[BaseNodePostprocessor] or None	No	Post-retrieval processors (re-rankers, filters, etc.)
response_mode	ResponseMode	No (default: COMPACT)	Strategy for synthesizing responses from retrieved nodes
text_qa_template	BasePromptTemplate or None	No	Custom prompt template for QA synthesis
refine_template	BasePromptTemplate or None	No	Custom prompt template for refine iterations
summary_template	BasePromptTemplate or None	No	Custom prompt template for tree summarization
output_cls	BaseModel or None	No	Pydantic model for structured output parsing
use_async	bool	No (default: False)	Whether to use async LLM calls during synthesis
streaming	bool	No (default: False)	Whether to enable streaming token output
verbose	bool	No (default: False)	Whether to print intermediate synthesis steps

Outputs

Name	Type	Description
query_engine	BaseQueryEngine	Fully configured query engine ready for .query() calls

Usage Examples

Basic Query Engine

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader

documents = SimpleDirectoryReader("data").load_data()
index = VectorStoreIndex.from_documents(documents)

# Create query engine with default settings (compact mode)
query_engine = index.as_query_engine()

Custom Response Mode and LLM

from llama_index.core import VectorStoreIndex
from llama_index.llms.openai import OpenAI

index = VectorStoreIndex.from_documents(documents)

# Use refine mode for thorough multi-pass synthesis
query_engine = index.as_query_engine(
    llm=OpenAI(model="gpt-4", temperature=0),
    response_mode="refine",
    similarity_top_k=5,
    streaming=True,
)

With Node Postprocessors

from llama_index.core import VectorStoreIndex
from llama_index.core.postprocessor import SimilarityPostprocessor

index = VectorStoreIndex.from_documents(documents)

# Filter out low-similarity nodes before synthesis
query_engine = index.as_query_engine(
    response_mode="tree_summarize",
    node_postprocessors=[
        SimilarityPostprocessor(similarity_cutoff=0.7),
    ],
)

Related Pages

Implements Principle

• Principle:Run_llama_Llama_index_Query_Engine_Creation